Post on 02-Jun-2018
transcript
8/11/2019 financial crisis 2011 (1).pdf
1/16
Liquidity risk management and credit supply in the financial crisis$
Marcia Millon Cornett a, Jamie John McNutt b, Philip E. Strahan c, Hassan Tehranian d,n
a Bentley University, USAb Southern Illinois University, Carbondale, USAc Boston College and NBER, USAd Boston College, USA
a r t i c l e i n f o
Article history:
Received 7 May 2010
Received in revised form
5 October 2010
Accepted 2 November 2010Available online 8 March 2011
JEL classification:
G01
G11
G21
Keywords:
Financial institutions
Liquidity riskFinancial crisis
a b s t r a c t
Liquidity dried up during the financial crisis of 20072009. Banks that relied more
heavily on core deposit and equity capital financing, which are stable sources of
financing, continued to lend relative to other banks. Banks that held more illiquid assets
on their balance sheets, in contrast, increased asset liquidity and reduced lending. Off-
balance sheet liquidity risk materialized on the balance sheet and constrained new
credit origination as increased takedown demand displaced lending capacity. We
conclude that efforts to manage the liquidity crisis by banks led to a decline in credit
supply.
& 2011 Elsevier B.V. All rights reserved.
1. Introduction
In this paper, we study how banks managed the
liquidity shock that occurred during the financial crisis
of 20072009 by adjusting their holdings of cash and
other liquid assets, as well as how these efforts to weather
the storm affected credit availability. Because the Federal
Reserve sets the aggregate supply of liquidity in the
banking system, focusing on only time series variation
in liquidity merely illustrates choices made by the Fed
(that is, the aggregate supply of liquidity). Our strategy
instead is to put a spotlight on within-bank variation in
holdings of cash and other liquid assets, which allows for
an understanding of why some banks chose to build up
liquidity faster than others during the crisis. This
approach helps explain why the Feds efforts to stimulate
the economy with traditional tools of monetary policy
were ineffective.
Our empirical model starts with the premise that
banks hold cash and other liquid assets as part of their
overall strategy to manage liquidity risk. In modern
banks, liquidity risk stems more from exposure to
undrawn loan commitments, the withdrawal of funds
from wholesale deposits, and the loss of other sources of
short-term financing than from the loss of demand
deposits (e.g., Diamond and Dybvig, 1983). With both
explicit and implicit government backing, deposits are
unlikely to leave the banking system during crises. For
example,Gatev and Strahan (2006)find inflows of depos-
its during periods of low market liquidity, while
Contents lists available at ScienceDirect
journal homepage: www.elsevier.com/locate/jfec
Journal of Financial Economics
0304-405X/$ - see front matter & 2011 Elsevier B.V. All rights reserved.doi:10.1016/j.jfineco.2011.03.001
$We are grateful to Effi Benmelech (the referee), G. William Schwert
(the editor), Robert DeYoung, Jason T. Greene, David Rakowski, and
seminar participants at Boston College, Boston University, Dartmouth
University, the Federal Reserve Board of Governors and the Federal
Reserve Banks of New York and Kansas City, Harvard University, South-
ern Illinois UniversityCarbondale, and the Deutsche Bundesbank,
European Banking Center, and European Business Schools Joint Con-
ference on Liquidity and Liquidity Risk for their helpful comments.n Corresponding author.
E-mail address: hassan.tehranian@bc.edu (H. Tehranian).
Journal of Financial Economics 101 (2011) 297312
http://-/?-http://www.elsevier.com/locate/jfechttp://localhost/var/www/apps/conversion/tmp/scratch_10/dx.doi.org/10.1016/j.jfineco.2011.03.001mailto:hassan.tehranian@bc.eduhttp://localhost/var/www/apps/conversion/tmp/scratch_10/dx.doi.org/10.1016/j.jfineco.2011.03.001http://localhost/var/www/apps/conversion/tmp/scratch_10/dx.doi.org/10.1016/j.jfineco.2011.03.001mailto:hassan.tehranian@bc.eduhttp://localhost/var/www/apps/conversion/tmp/scratch_10/dx.doi.org/10.1016/j.jfineco.2011.03.001http://www.elsevier.com/locate/jfechttp://-/?-8/11/2019 financial crisis 2011 (1).pdf
2/16
Pennacchi (2009)does not find such flows during the pre-
Federal Deposit Insurance Corporation (FDIC) period.
Together this suggests that deposits insulate banks fromliquidity risk due to the advent of government guarantees.
Liquidity risk from loan commitments, for example, was
evident in aggregate data when the commercial paper
markets froze following the September 2008 failure of
Lehman Brothers. Issuers responded by taking down
funds from commercial paper backup lines issued by
banks, leading to a decline in commercial paper out-
standing and an increase in bank lending (Fig. 1). At the
same time, banks lost wholesale funds but gained retail
deposits (Fig. 2).1 We show that banks more exposed to
this liquidity risk increased their holdings of liquid assets,
which in turn reduced their capacity to make new loans.
On the asset side of balance sheets, banks holdingassets with low market liquidity expanded their cash
buffers during the crisis. Specifically, banks that held
more loans, mortgage-backed securities (MBS), and
asset-backed securities (ABS) tended to increase holdings
of liquid assets and decrease investments in loans and
new commitments to lend. Because of concerns about the
liquidity of loans and securitized assets, these banks
rationally protected themselves by hoarding liquidity, to
the detriment of their customers and markets. Turning to
70
80
90
100
110
120
130
June2007
=
100
Commercial Paper Bank C&I Loans
Fig. 1. Business lending rises as commercial paper moves back on the balance sheet. This figure shows the growth of commercial paper and bank business loans
outstanding from June 2007 through November 2008. Data are obtained from the website of the Board of Governors of the Federal Reserve (www.chicagofed.org).
-15%
-10%
-5%
0%
5%
10%
15%
Wholesale Deposits Core Deposits
Fig. 2. Growth in deposits. This figure shows the weekly percentage change in core and wholesale deposits at commercial banks from September 10,
2008 through December 31, 2008. Core deposits include transactions deposits plus fully insured (o$100,000) time deposits. Wholesale deposits include
time deposits over $100,000. Data are obtained from the Federal Reserves H8 weekly data on bank assets and liabilities.
1 Gorton (2009) and Gorton and Metrick (2009) draw parallelsbetween the increase in haircuts in the repo markets and banking
(footnote continued)
panics and bank runs. These effects were greatest at large nonbankfinancial institutions.
M.M. Cornett et al. / Journal of Financial Economics 101 (2011) 297312298
http://www.chicagofed.org/http://www.chicagofed.org/8/11/2019 financial crisis 2011 (1).pdf
3/16
the right-hand side of the balance sheet, banks with stable
sources of financing were less constrained by the crisis
and, thus, were able to continue to lend. Banks using more
core deposits (all transactions deposits plus other insured
deposits) and more equity capital to finance their assets
saw significant increases in lending, relative to banks that
relied more on wholesale sources of debt financing. The
results hold when we control for aggregate time effects,bank fixed effects, measures of loan demand, and the
effects of financial structure during normal market con-
ditions. Moreover, the results are consistent across both
large and small bank samples, although the economic
impact is generally bigger for the large bank sample.
We also test how banks managed shocks to loan
demand stemming from preexisting unused loan commit-
ments (held off the balance sheet). Unused commitments
expose banks to liquidity risk, which became manifest
when takedown demand increased following the collapse
of Lehman Brothers. We find that banks with higher levels
of unused commitments increased their holdings of liquid
assets (i.e., their precautionary demand for liquidityincreased) and also cut back on new credit origination
(measured by summing on-balance sheet loans with off-
balance sheet loan commitments). Loan commitment
drawdowns thus displaced new credit origination during
the crisis.
Our paper extends in three ways the empirical analysis
ofIvashina and Scharfstein (2010), who use Dealscan data
to show that new bank lending growth fell less at banks
funded with deposits and more at banks exposed to
unused credit lines. First, we show that liquidity risk
exposure is not only negatively correlated with loan
growth in the crisis, but it is also positively correlated
with the growth in liquid assets. These parallel resultssupport the interpretation that efforts to build up balance
sheet liquidity displaced funding to support new lending.
Second, we have a much larger and richer data set [drawn
from the quarterly Federal Financial Institutions Exam-
ination Council (FFIEC) Reports of Income and Condition
(Call Reports)], which allows us to explore more dimen-
sions of liquidity risk exposure and to quantify implica-
tions of our results for overall credit supply. For example,
we show that the market liquidity of bank assets nega-
tively affected their accumulation of liquid assets and
positively affected their loan growth. Also, we show that it
is core deposits, not total deposits, which provided stable
funding to banks. Third, we work to rule out loan demandexplanations for our results by exploiting geographical
exposure from the FDIC Summary of Deposits and loan
account data available from Call Reports.
Because we look at the whole banking system, our
regressions can help draw out the macroeconomic impli-
cations of our results. We quantify how much credit
would have contracted if banks had entered the fall of
2008 less exposed to liquidity risk. This analysis suggests
that the pressure on bank balance sheets from takedowns
on preexisting loan commitments and funding problems
from wholesale markets account for most of the decline in
new credit production. New credit productionthat is,
the sum of both on-balance sheet loans and undrawncommitmentsfell by about $500 billion in the fourth
quarter of 2008 (out of a total of slightly more than $14
trillion of total loans plus undrawn commitments to lend
at the end of 2008). Had liquidity exposure been in the
lower quartile across the whole banking system, our
estimates suggest that new credit would have fallen by
just $87 billion, or almost 90% less than the unadjusted
figure.
In the remainder of the paper, we provide in Section 2a brief chronology of the financial crisis to justify our
identification strategy based on time variation of the TED
spread as a measure of liquidity strains on the banking
system. After laying out the drivers of bank liquidity risk
to motivate our empirical model, we describe the data
and results in Section 3. We conclude in Section 4.
2. The TED spread during the financial crisis of 2007
2009
The financial crisis of 20072009 is the biggest shock
to the US and worldwide financial system since the 1930s
and offers a unique challenge to both financial institu-tions and regulators understanding of liquidity produc-
tion and liquidity risk management.2 Fig. 3illustrates the
time series of new loan originations to large businesses
from Loan Pricing Corporations Dealscan database from
2000 to the end of 2008. During the 20012002 recession,
both lines of credit and term loans declined as would be
expected during a mild recession. But, this earlier decline
pales relative to the steep drop in new lending beginning
in the middle of 2007.
The crisis began in the summer of 2007 when the
asset-backed commercial paper market began to unravel
in the face of uncertainty about the value and liquidity of
some mortgage-backed securities (Acharya and Schnabl,2010). The brewing crisis can be seen in the TED spread
[the difference between the three-month London Inter-
bank Offered Rate (LIBOR) and the three-month Treasury
rate], which spiked above 200 basis points. From then
until the spring of 2009, the TED spread (as well as other
similar indicators) remained both elevated and volatile.
The TED spread is an indicator of perceived credit risk in
the general economy. This is because T-bills are consid-
ered risk-free, while LIBOR reflects the credit risk of
lending to commercial banks. An increase in the TED
spread indicates that lenders believe the risk of default
on interbank loans (i.e., counterparty risk) is increasing.
We plot the time series variation of the TED spread fromthe beginning of 2006 to the end of the second quarter of
2009 inFig. 4.[Fig. 4also shows (in the shaded area) the
period we designate as the crisis period in our robustness
test below.]
Time variation in the TED spread tracks the severity of
the crisis closely. For instance, the TED spread spiked in
March 2008 as Bear Stearns failed. Conditions improved
following the Bear Stearns bailout, and the TED spread
subsided. In the summer of 2008, however, concerns
about mortgage foreclosures rose, further downgrades of
mortgage-backed securities by the credit rating agencies
2 SeeBrunnermeier (2009)for a detailed discussion of these events.
M.M. Cornett et al. / Journal of Financial Economics 101 (2011) 297312 299
8/11/2019 financial crisis 2011 (1).pdf
4/16
occurred, and losses to holders of these securities
mounted. Losses on mortgages and mortgage-backedsecurities eventually led to the failure of several financial
institutions, notably, Fannie Mae (Federal National Mort-
gage Association) and Freddie Mac (Federal Home Loan
Mortgage Corporation) and then American International
Group, Inc. (AIG) and Lehman Brothers. The depth of the
crisis dramatically expanded when financial markets
were shocked by the collapse of these institutions, along
with the distressed sale of Merrill Lynch to Bank of
America. The panic soon spread, leading to the expansion
of insurance on deposits and interbank funds, first in
Europe and then very quickly in the United States. The
crisis truly abated only in the spring of 2009 when the
stress tests of the large US banks brought private capitalback into the system.
3. Empirical strategy and results
In this section, we first discuss the determinants of
bank liquidity risk and then describe our empirical model,
data, and results.
3.1. Liquidity risk management
Liquidity production is central to all theories of finan-
cial intermediation. First, asymmetric information proces-
sing allows banks to create liquidity through their asset
transformation function (seeDiamond and Dybvig, 1983).
Second, banks provide liquidity to borrowers in the
form of credit lines and to depositors by making funds
available on demand. These functions leave banks vulner-able to systemic increases in demand for liquidity from
0
200
400
600
800
1,000
1,200
Billions
ofdollars
New Term Loans New Credit Lines
Fig. 3. Business loan originations collapse. This figure shows the dollar value of new term loans and credit lines issued to large businesses from 2000
(before the financial crisis) through 2008 (at the height of the financial crisis). Data used to construct the figure are obtained from the Loan Pricing
Corporations Dealscan database.
0.0%
0.5%
1.0%
1.5%
2.0%
2.5%
3.0%
3.5%
4.0%
4.5%
5.0%
Fig. 4. The TED spread. This figure shows movements in the TED spread from 2006 through the second quarter of 2009. The TED spread is calculated as
the difference between the three-month London Interbank Offered Rate (LIBOR) rate [obtained from the website of the Bulgarian National Bank (www.
bnb.bg/#)] and the three-month Treasury rate [from the Federal Reserve Economic Data (FRED) website of the Federal Reserve Bank of St. Louis ( http://
research.stlouisfed.org/fred2/)]. The shaded area includes the period we designate the crisis period in our empirical analysis.
M.M. Cornett et al. / Journal of Financial Economics 101 (2011) 297312300
http://www.bnb.bg/#http://www.bnb.bg/#http://research.stlouisfed.org/fred2/http://research.stlouisfed.org/fred2/http://research.stlouisfed.org/fred2/http://research.stlouisfed.org/fred2/http://www.bnb.bg/#http://www.bnb.bg/#8/11/2019 financial crisis 2011 (1).pdf
5/16
borrowers and, at the extreme, can result in runs on banks
by depositors. In the traditional framework of banking,
runs can be prevented, or at least mitigated, by insuring
deposits and by requiring banks to issue equity and to
hold cash reserves (e.g., Diamond and Dybvig, 1983;
Gorton and Pennacchi, 1990). Systemic increases in
demand for liquidity from borrowers, in contrast, depend
on external market conditions and thus are harder forindividual banks to manage internally. For example, when
the supply of overall market liquidity falls, borrowers turn
to banks en masse to draw funds from existing credit lines
(Gatev and Strahan, 2006).
Diamond and Rajan (2001b) note that while banks
provide liquidity to borrowers, the loans themselves are
relatively illiquid assets for banks. Subsequently, when
banks require liquidity, they could sell the loans (e.g., sell
and securitize mortgages to create mortgage-backed
securities) or use the loans as collateral (e.g., mortgages
serve as collateral for mortgage-backed bonds issued by
the banks) (see Bhattacharya and Thakor (1993);
Diamond and Rajan (2001b)). Such sales, however,become more difficult when market liquidity becomes
scarce. Thus, Diamond and Rajan (2001b) also note that
banks can ration credit if future liquidity needs are likely
to be high.Diamond and Rajan (2001a)suggest banks can
be fragile because they must provide liquidity to deposi-
tors on demand and because they hold illiquid loans.
Further, demands by depositors can occur at undesirable
times, i.e., when loan payments are uncertain and when
there are negative aggregate liquidity shocks. In addition,
Kashyap, Rajan, and Stein (2002) note similarities
between some off-balance sheet (i.e., contingent) assets
and on-balance sheet assets. In particular, an off-balance
sheet loan commitment becomes an on-balance sheetloan when the borrower chooses to draw on the commit-
ment.Berger and Bouwman (2009)find that roughly half
of the liquidity creation at commercial banks occurs
through these off-balance sheet commitments. Thus,
banks stand ready to supply liquidity to both borrowers
and insured retail depositors and can enjoy synergies
when depositors fund loan commitments. Recent evi-
dence lends support to this notion. Gatev, Schuermann,
and Strahan (2009)find deposits effectively hedge liquid-
ity risk inherent in unused loan commitments and the
effect is more pronounced during periods of tight
liquidity.
The role of bank equity capital also plays a part in theliquidity provision function of commercial banks.Diamond
and Rajan (2000)suggest equity capital can act as a buffer
to protect depositors in times of distress. However, holding
excessive equity capital can reduce liquidity creation and
the flow of credit.Gorton and Winton (2000)conclude that
regulators should be especially aware of these effects
during recessionary environments, i.e., periods when reg-
ulators could want to increase capital standards to reduce
the threat of bank failures. Recent evidence suggests bank
size can affect which effect dominates. Berger and
Bouwman (2010) find that higher capital levels crowd
out depositors and decrease liquidity creation at smaller
banks, but higher capital levels absorb risk and increaseliquidity creation at larger banks.
Banks facilitate their operations with more than retail
deposits and equity capital, most notably with uninsured
wholesale deposits and subordinated notes and deben-
tures. Researchers and regulators have long been inter-
ested in these alternate funding mechanisms and their
role in imparting market discipline on bank behavior.3 For
example, Hannan and Hanweck (1988) find uninsured
depositors require higher interest rates at riskier banks,and Maechler and McDill (2006) suggest uninsured
depositors might not supply liquidity to weak banks at
any price. Avery, Belton, and Goldberg (1988) find little
evidence that holders of bank-issued subordinated notes
and debentures effectively constrain bank risk. However,
restrictive covenants have been found to be more com-
mon in debt contracts when banks are riskier (seeGoyal,
2005;Ashcraft, 2008).
Size also matters. That is, the markets perception of
the risk of a bank can depend on the size of the bank. The
Comptroller of the Currencys statement before Congress
on September 19, 1984 that some financial institutions
are too-big-to-fail (TBTF) was a positive wealth event forbanks deemed TBTF (seeOHara and Shaw, 1990). Further
evidence is provided by Black, Collins, Robinson, and
Schweitzer (1997), who observe a flight to quality as
evidenced by changes in institutional ownership of TBTF
bank equity shares.
3.2. Empirical specification
The discussion above suggests four key drivers of
liquidity risk management for banks: (1) the composition
of the asset portfolio (i.e., the market liquidity of assets),
(2) core deposits as a fraction of total financial structure,(3) equity capital as a fraction of financial structure, and
(4) funding liquidity exposure stemming from loan com-
mitments (i.e., new loan originations via drawdowns).
Asset size also likely relates to liquidity management, but
it proxies for many other sources of heterogeneity. Hence,
we include this variable in all of our regressions but
refrain from interpreting its effect.
Our identification strategy is based on the premise that
tight liquidity conditions during the financial crisis, mea-
sured by the TED spread, surprised banks and thus
changed their management of liquidity risk exposure.
That is, banks with high liquidity risk exposure would
be expected to build up cash and other liquid assets andalso to reduce new lending (particularly new commit-
ments to lend) more than banks with low liquidity risk
exposure when the TED spread spikes. We test this idea
by interacting the TED spread with our four measures of
liquidity exposure.
We build a quarterly panel data set from the beginning
of 2006 through the second quarter of 2009 that includes
all commercial banks as described below. This sample has
observations before and during the financial crisis, at least
judging by movements in TED spreads. With the panel
3 See Flannery (1998) for an overview of the role of market
discipline as it relates to regulatory supervision and Flannery (2001)for an overview of the notion of market discipline.
M.M. Cornett et al. / Journal of Financial Economics 101 (2011) 297312 301
8/11/2019 financial crisis 2011 (1).pdf
6/16
approach we can sweep out aggregate trends, such as the
Feds expansion of the supply of overall liquidity, as well
as bank fixed effects to account for unobserved hetero-
geneity. Moreover, we can control for the normal impact
(or correlation) of the liquidity exposure measures in our
model and focus on the interaction of the TED spread with
those variables. To be specific, we estimate the following
three regressions:
DLiquid Assetsi,t=Assetsi,t1
T1t B1i b
1Illiquid Assets=Assetsi,t1
b2
Illiquid Assets=Assetsi,t1TEDt
b3
Core Deposits=Assetsi,t1
b4
Core Deposits=Assetsi,t1TEDt
b5
Capital=Assetsi,t1
b6
Capital=Assetsi,t1TEDt
b7
Commit=CommitAssetsi,t1
b8
Commit=CommitAssetsi,t1TEDt
b9Log Assetsi,t1
b10
Log Assetsi,t1TEDtei,t, 1
DLoansi,t=Assetsi,t1
T2t B2i g
1Illiquid Assets=Assetsi,t1
g2Illiquid Assets=Assetsi,t1TEDt
g3Core Deposits=Assetsi,t1g4Core Deposits=Assetsi,t1TEDt
g5Capital=Assetsi,t1g6Capital=Assetsi,t1TEDt
g7Commit=CommitAssetsi,t1
g
8Commit=
Commit
Assetsi,t1
TEDt
g9Log Assetsi,t1g10Log Assetsi,t1TEDtZi,t, 2
and
DCrediti,t=CommitAssetsi,t1
T3t B3i l
1Illiquid Assets=Assetsi,t1
l2
Illiquid Assets=Assetsi,t1TEDt
l3
Core Deposits=Assetsi,t1
l4
Core Deposits=Assetsi,t1TEDt
l5
Capital=Assetsi,t1
l6
Capital=Assetsi,t1TEDt
l7Commit=CommitAssetsi,t1
l8
Commit=CommitAssetsi,t1TEDt
l9
Log Assetsi,t1
l10
Log Assetsi,t1TEDtmi,t, 3
where T1, T2, and T3 are time effects that sweep out
aggregate shocks and B1, B2, and B3 are bank-level fixed
effects that absorb unobserved heterogeneity at the bank
level. Because our panel covers only three and a half
years, we feel that the assumption that bank effects are
fixed over time is reasonable. In constructing standard
errors, we consistently cluster errors at the bank level to
account for potential serial correlation at the bank level.Also, because we normalize all financing variables by total
assets in the three regressions, the coefficients on these
variables (i.e., Core Deposits/Assets and Capital/Assets)
represent the effect of moving funding from capital (or
deposits) to the omitted category (mostly wholesale
sources of short-term debt). In other words, these coeffi-
cients can be interpreted only relative to the omitted
category. We estimate each of these relations separately
for large (4$1 billion in assets) and small (r$1 billion inassets) banks. Regression variables are defined and their
descriptive statistics are discussed in detail in Section 3.3.
Variables are winsorized at the 1st and 99th percentiles.
Regression Eq. (1) tests how banks adjust their hold-
ings of liquid assets, regression Eq. (2) tests how bank
lending on the balance sheet adjusts, and regression
Eq. (3) tests how total credit origination adjusts. Loans
on the balance sheet vary both because banks expand
new (net) lending and because borrowers draw funds
from preexisting commitments (off-balance sheet items
while undrawn). Hence, takedowns of previous commit-
ments, which increased during the financial crisis after
the commercial paper market dried up, could displacelending capacity in the banking system. To take account of
these movements from off-balance sheet to on-balance
sheet items, we construct a variableCreditfor regression
Eq. (3), equal to the sum of loans on the balance sheet
plus undrawn loan commitments off the balance sheet.
Thus, results from this regression reflect increases in bank
credit from new originations of both loans and loan
commitments. That is, loan commitment drawdowns do
not affect this measure of overall credit supply because
unused commitments decrease by the same level that
loans increase. Such an interpretation is not possible by
looking only at changes in loans reported on the balance
sheet. For this specification, we normalize the dependentvariable by total loan commitments plus total assets
instead of just total assets.
During the crisis, banks were no longer able to secur-
itize loans (originate and distribute) to the extent they
had prior to the crisis. Further, market liquidity for
mortgage-backed securities and asset-backed securities
became all but nonexistent. Accordingly, we expect banks
that held more of these illiquid assets during the crisis
period to increase their holdings of liquid assets and
constrain new lending and credit creation. Thus, we
expect b240, g2o0, and l2o0. If core deposits andcapital act as stable sources of financing during the crisis,
then we expect banks with higher levels of both to bemore willing to run down their liquidity buffers. That is,
b4o0 andb6o0. Further, if these stable sources of funds
allowed banks to continue to lend during the crisis, we
expectg440 andg640 (andl440 andl640). The effectof unused loan commitments is harder to sign ex ante
because banks with greater unused commitments are
exposed to liquidity risk (suggesting b840) but also
experience a greater increase in loan demand in the crisis
(so,g840 as well). However, we would expect banks withgreater exposure to liquidity risk from lending via com-
mitments to reduce total credit originations (so, l8o0).
In addition to the models in regression Eqs. (1)(3), we
report models using an indicator variable for the crisisperiod instead of the TED spread. We set the crisis
M.M. Cornett et al. / Journal of Financial Economics 101 (2011) 297312302
8/11/2019 financial crisis 2011 (1).pdf
7/16
indicator equal to one from 2007Q3 through 2009Q2 (see
Fig. 4).
Our strategy exploits the exogenous shock to overall
liquidity as measured by the TED spread. Hence, we do
not attempt to interpret the direct effects of the variables
in regression Eqs. (1)(3). Said differently, we are side-
stepping the problem that policy makers, the Fed in this
case, chose to increase aggregate liquidity. As is wellknown, the Fed expanded its balance sheet from about
$800 billion to a little more than $2 trillion during the
fourth quarter of 2008, leading to an increase in cash in
the banking system. Instead, regression Eq. (1) allows us
to understand how that liquidity was distributed across
the banking system, which is endogenously determined
by variations in banks liquidity demands.
3.3. Data
We build our panel data set from the quarterly FFIEC
Call Reports, which all regulated commercial banks
file with their primary regulator.4 Because some banks
are owned by a common holding company, we aggregate
the bank-level data for banks with common ownership
because these ownership ties could foster liquidity
sharing across subsidiaries (see Houston, James, and
Marcus, 1997). Specifically, we sum Call Reports data at
the highest holding company level for multibank holding
companies.
Call Reports contain detailed on- and off-balance sheet
information for all banks. Specific to our study, we collect
information on bank assets, deposits, capital, and off-
balance sheet, undrawn loan commitments. Following
Federal Deposit Insurance Corporation and Federal
Reserve guidelines, we segregate banks into two size
groups based on beginning of quarter book value of
assets: large banks have assets of greater than $1 billion
and small banks have assets less than or equal to $1
billion. Banks with asset growth greater than 10% during a
quarter are dropped during that quarter to mitigate the
effect of large mergers on changes in liquid assets, loans,
and credit supply. Table 1 lists the distribution of the
sample banks by quarter.
From Call Reports data we build the dependent vari-
ables for our three regression models: change in liquid
assets during the quarter divided by beginning of quarter
total assets (DLiquid Assetsi,t/Assetsi,t1), where liquid
assets includes cash plus non-asset-backed securities;
change in loans during the quarter divided by beginning
of period assets (DLoansi,t/Assetsi,t1); and change in the
sum of loans plus undrawn commitments divided by the
sum of total assets plus undrawn commitments at the
beginning of the quarter (DCrediti,t/(CommitAssets)i,t1).
While Loutskina (in press) finds that securitizable assets
offer banks a liquidity buffer during normal markets,
these markets dried up starting in the summer of 2007.
We thus drop mortgage-backed securities and asset-
backed securities from our definition of liquid assets and
instead include them in our measure of illiquid assets.5 In
addition, we report an alternative measure of the change
in liquid assets that includes just cash plus US govern-
ment securities plus Fed funds sold and securities pur-
chased under agreements to resell, but it leaves out other
securities. US government securities clearly maintained
their liquidity in the crisis, whereas Fed funds and reverse
repos remained highly liquid due to their very short
maturity.6
Explanatory variables in the regressions include thefraction of the firms investment portfolio of assets that
are illiquid at the beginning of the period (Illiquid Assets/
Assetsi,t1), the fraction of the firms balance sheet
financed with core deposits at the beginning of the period
(Core Deposits/Assetsi,t1), the fraction of the balance sheet
(risk-weighted assets) financed by Tier 1 capital at the
beginning of the period (common stockholders equity
plus qualifying perpetual preferred stock) (Capital/
Assetsi,t1), the ratio of unused commitments to commit-
ments plus assets at the beginning of the period (Commit/
(CommitAssets)i,t1), and the log of total assets at the
beginning of the period (Log Assetsi,t1).7 Each of these
variables is included in the regressions independently and
Table 1
Number of commercial banks examined between 2006Q1 and 2009Q2.
This table lists the distribution of the sample banks by quarter. We
segregate banks into two size groups based on beginning of quarter book
value of assets. Large banks are those banks with beginning of quarter
assets greater than $1 billion, and small banks are those banks with
beginning of quarter assets less than $1 billion. Bank asset size is
collected from Federal Financial Institutions Examination Council Call
Reports of Condition and Income found on the website of the Federal
Reserve Bank of Chicago (www.chicagofed.org).
Year Quarter Large Small Total
2006 1 430 5,784 6,214
2 432 5,789 6,221
3 442 5,731 6,173
4 434 5,563 5,997
2007 1 455 5,664 6,119
2 453 5,670 6,123
3 457 5,639 6,096
4 462 5,539 6,001
2008 1 459 5,460 5,919
2 483 5,583 6,066
3 485 5,575 6,060
4 468 5,440 5,908
2009 1 491 5,506 5,997
2 511 5,638 6,149
4 Call Reports data are publicly available at the website of theFederal Reserve Bank of Chicago (www.chicagofed.org).
5 Specifically, Liquid Assetsnoninterest-bearing cash balances
interest-bearing cash balances non-MBS and non-ABS held-to-matur-
ity (HTM) securitiesnon-MBS and non-ABS available-for-sale (AFS)
securitiesfed funds sold securities purchased under agreements to
resell.6 Repurchases agreements (that is, the sale of assets with agreement
to repurchase them) were a source of funding problems for firms such as
Bear Stearns and Lehman Brothers. However, we are looking at banks as
lenders, not as borrowers, in this market. That is, our measure of liquid
assets includes only so-called reverse repos, in which the position acts as
an asset to the bank instead of as a liability.7 Specifically, Illiquid Assetsloans and leases net of unearned
income and allowances MBS and ABS HTM securities MBS and ABS
AFS securities. Core deposits are defined as the sum of deposits under$100,000 plus all transactions deposits.
M.M. Cornett et al. / Journal of Financial Economics 101 (2011) 297312 303
http://www.chicagofed.org/http://www.chicagofed.org/http://www.chicagofed.org/http://www.chicagofed.org/http://www.chicagofed.org/http://www.chicagofed.org/http://www.chicagofed.org/8/11/2019 financial crisis 2011 (1).pdf
8/16
8/11/2019 financial crisis 2011 (1).pdf
9/16
total credit are both lower in the crisis quarters relative to
the noncrisis quarters. For large banks, the mean (median)
percentage change in loans to assets is 1.39 (1.33) during
the noncrisis period and 0.85 (0.83) during the crisis
period; the mean (median) percentage change in credit
supply to assets plus commitments is 1.60 (1.46) during
the noncrisis period and 0.50 (0.54) during the crisis
period. For small banks, the mean (median) percentagechange in loans to assets is 1.32 (1.06) during the
noncrisis period and 1.04 (0.76) during the crisis period;
the mean (median) percentage change in credit supply to
assets plus commitments is 1.41 (1.06) during the non-
crisis period and 0.96 (0.65) during the crisis period.
The differences are generally even more pronounced if
just the fourth quarter of 2008 (during the height of the
financial crisis) values are compared with those of the
noncrisis period. While mean and median liquid assets
fall on average during both noncrisis and crisis quarters,
stores of liquidity increase during the fourth quarter of
2008 when the Fed engineered the massive expansion of
overall liquidity supply. For large banks, the mean (med-ian) percentage change in liquid assets to assets is 0.18
( 0.09) during the noncrisis period and 0.22 ( 0.20)
during the crisis period, yet it is 0.54 (0.15) in 2008Q4
(the peak crisis quarter). For small banks, the mean
(median) percentage change in liquidity to assets is
0.34 ( 0.21) in the crisis period, but 0.16 (0.16) in
2008Q4. We see the same pattern with the alternative
measure of liquidity (change in cash plus US government
securities plus Fed funds sold and securities purchased
under agreements to resell to assets). We also test equal-
ity of the means and medians across the two samples. We
find that while credit grew faster at large banks during
normal quarters, it grew significantly slower during thecrisis quarters. This likely reflects the greater effect of the
liquidity crisis on the larger banks.
Panels G and H ofTable 2 list summary statistics for
large and small banks, respectively, on the independent
variables used in the regression analysis. Comparing
characteristics, Table 2 shows that small banks tend to
rely more on core deposits and capital to finance their
balance sheets than large banks. Core deposits to assets
and capital to assets at small banks are, on average,
66.44% and 18.56%, respectively, and at large banks they
are 58.75% and 11.69%, respectively. Further, large banks
have more illiquid assets per dollar of total assets than
small banks (77.80% versus 70.27%) and also hold agreater fraction of unused commitments compared with
small banks (16.79% versus 9.18%). All of these differences
are statistically significant at the 1% level. These simple
comparisons suggest that large banks are more exposed
to liquidity risk than small banks across all four dimen-
sions: more undrawn commitments, less capital, less
reliance on core deposits, and lower liquidity of balance
sheet assets.
3.4. Regression results
Table 3reports our models for regression Eqs. (1)(3).
Panel A reports the regressions for large banks (over $1billion in beginning-of-quarter assets) and Panel B reports
the regressions for small banks. A consistent pattern
emerges: During the crisis, liquidity risk exposure led to
greater increases in liquid assets, mirrored by greater
decreases in credit origination. The interaction between
the TED spread and each exposure measure enters the
regressions in every case with opposite signs (compare
Columns 1 and 4 in Table 3). For example, Illiquid Assets/
AssetsnTED enters the liquid asset growth equation posi-tively (2.423, Column 1) and the credit growth equation
negatively ( 1.340, Column 4). The same holds for Core
Deposits/AssetsnTED, Capital/AssetsnTED, and Commit/
(CommitAssets)nTED.
The pattern holds for both large and small banks (Panels
A and B ofTable 3). Taken together, this is strong evidence
that banks built up liquidity buffers to offset the increased
risk during the crisis and, as a result, had to cut back on
credit production. Liquidity risk management thus helps
explain changes in credit supply across banks. The results
for loan growth (Column 3) are consistent with those for
total credit production (Column 4) across three of the four
liquidity variables, the exception being unused commit-ments. For this variable, we observe a positive effect of
Commit/(CommitAssets)nTED, reflecting the increased
takedown demand during the crisis in the loan growth
equation as funds moved from off-balance sheet accounts
to on-balance sheet accounts. This occurs despite a rela-
tively larger drop in total credit production for banks that
were more exposed to preexisting commitments.
InTable 4, we replace the TED spread with an indicator
variable equal to one during the quarters in which TED was
elevated, i.e., 2007Q3 through 2009Q2. This approach has
the advantage of better robustness because the indicator is
by construction free of outliers. However, this indicator has
a drawback in that it misses the activity during the keyfourth quarter of 2008 when markets dried up spectacu-
larly following the Lehman bankruptcy and AIG bailout.
The results are consistent in terms of sign patterns with
those inTable 3. Magnitudes appear different because the
quarterly average of the daily TED spread varies from 37
basis points (in 2006Q1) to 250 basis points (in 2008Q4),
while the indicator varies between zero and one.
One of our most consistent findings is that core deposits
(transactions deposits plus other insured funds) helped
banks sustain lending. In fact, in unreported tests we add
wholesale deposits (uninsured, nontransactions deposits)
as an explanatory variable but find that these do not
correlate positively with credit production. While deposi-tors can withdraw transaction deposits on demand, they
rarely do. Thus, banks use these deposits to fund loans and
commitments. They act as a substitute for liquid assets.
Diamond and Dybvigs model of asset transformation
ties bank fragility to demandable deposits. In contrast to
this classic scenario, during the financial crisis funds were
leaving the securities markets and flowing into the bank-
ing system (the opposite of a run), and most of the funds
flowed into bank transactions deposit accounts.8 Further,
8 Billett, Garfinkel, and ONeal (1998) also find at the micro level
that distressed banks tend to substitute insured deposits for uninsureddeposits.
M.M. Cornett et al. / Journal of Financial Economics 101 (2011) 297312 305
8/11/2019 financial crisis 2011 (1).pdf
10/16
Table 3
Fixed effects regressions of liquid asset, loan, and credit supply growth on TED spread, firm characteristics, and interactions.
This table reports fixed effects regressions of quarterly growth in liquid assets standardized by beginning of period assets. The table also reports fixed
effects regressions of growth in loans standardized by beginning of period assets and growth in credit supply (i.e., growth in loans plus growth in unused
commitments) standardized by beginning of period assets plus unused commitments. The data are observed quarterly for a panel of US commercial
banks over the period 2006Q1 through 2009Q2. Large banks are those banks with beginning of quarter assets greater than $1 billion, and small banks are
those banks with beginning of quarter assets less than $1 billion. Commercial bank data, aggregated at the high holding company as appropriate, are from
Call Reports of Condition and Income accessed via the Federal Reserve Bank of Chicago website ( www.chicagofed.org). Banks with asset growth greater
than 10% during a quarter are dropped during that quarter to mitigate the effect of large mergers on changes in liquid assets, loans, and credit supply.
Variables are winsorized at the 1st and 99th percentiles. TED spread is the quarterly average of the daily difference between the three-month LondonInterbank Offered Rate (LIBOR) and the three-month US Treasury bill secondary market rate. LIBOR data are from the Bulgarian National Bank website
(www.bnb.bg/#) and Treasury bill data are from the Federal Reserve Economic Data (FRED) website of the Federal Reserve Bank of St. Louis ( http://
research.stlouisfed.org/fred2/). Standard errors, clustered at the firm level, are reported in parentheses. ***, **, and * denote that the coefficients are
statistically significantly different from zero at the 1%, 5%, and 10% level, respectively.
DLiquid assetst/
Assetst1
D(Cash US Treas Fed Funds Repost)/
Assetst1
DLoanst/
Assetst1
DCreditt/
(CommitAssets)t1
(1) (2) (3) (4)
Panel A: Large banks (total assets 4$1 billion)
Illiquid Assets/Assetst1 0.209*** 0.187*** 0.015 0.030*
(0.023) (0.020) (0.014) (0.017)
Illiquid Assets/Assetst1nTED 2.423*** 0.773 1.145* 1.340*
(0.863) (0.710) (0.678) (0.762)
Core Deposits/Assetst1 0.004 0.004 0.025*** 0.044***
(0.014) (0.010) (0.009) (0.011)
Core Deposits/Assetst1nTED 0.837 0.440 1.231*** 1.356**
(0.663) (0.601) (0.453) (0.555)
Capital/Assetst1 0.042 0.096** 0.042 0.088*
(0.050) (0.039) (0.037) (0.054)
Capital/Assetst1nTED 0.499 2.314 5.732** 5.984**
(2.233) (2.068) (2.637) (2.699)
Commit/(CommitAssets)t1 0.070*** 0.065*** 0.176*** 0.012
(0.023) (0.018) (0.027) (0.032)
Commit/(CommitAssets)
t1nTED
0.933 0.602 2.303** 2.753***
(0.961) (0.736) (0.937) (0.855)
Log Assetst1 0.016*** 0.017*** 0.017*** 0.027***
(0.005) (0.004) (0.005) (0.007)
Log Assetst1nTED 0.186*** 0.132** 0.064 0.131*
(0.072) (0.061) (0.052) (0.067)
Firm dummies Yes Yes Yes Yes
Quarterly time dummies Yes Yes Yes Yes
N 6,462 6,462 6,462 6,462
AdjustedR-squared 0.244 0.205 0.404 0.342
Panel B: Small banks (total assets r$1 billion)
Illiquid Assets/Assetst1 0.291*** 0.278*** 0.067*** 0.051***
(0.011) (0.007) (0.004) (0.006)
Illiquid Assets/Assetst1nTED 0.240 0.702*** 0.546*** 0.516***
(0.296) (0.265) (0.163) (0.183)
Core Deposits/Assetst1 0.028*** 0.039*** 0.002 0.002
(0.006) (0.005) (0.005) (0.006)
Core Deposits/Assetst1nTED 0.229 0.871*** 0.673*** 1.553***
(0.295) (0.268) (0.204) (0.238)
Capital/Assetst1 0.008 0.006 0.070*** 0.067***
(0.012) (0.013) (0.009) (0.012)Capital/Assetst1nTED 1.456*** 2.744*** 1.135*** 1.058***
(0.500) (0.471) (0.273) (0.325)
Commit/(CommitAssets)t1 0.167*** 0.152*** 0.295*** 0.137***
(0.010) (0.010) (0.009) (0.012)
Commit/(CommitAssets)
t1nTED
3.090*** 1.933*** 1.071** 3.034***
(0.611) (0.568) (0.459) (0.557)
Log Assetst1 0.040*** 0.039*** 0.015*** 0.028***
(0.003) (0.003) (0.004) (0.006)
Log Assetst1nTED 0.311*** 0.269*** 0.079*** 0.021
(0.033) (0.029) (0.022) (0.024)
Firm dummies Yes Yes Yes Yes
Quarterly time dummies Yes Yes Yes Yes
N 78,581 78,581 78,581 78,581
AdjustedR-squared 0.270 0.253 0.380 0.299
M.M. Cornett et al. / Journal of Financial Economics 101 (2011) 297312306
http://www.chicagofed.org/http://www.bnb.bg/#http://research.stlouisfed.org/fred2/http://research.stlouisfed.org/fred2/http://research.stlouisfed.org/fred2/http://research.stlouisfed.org/fred2/http://www.bnb.bg/#http://www.chicagofed.org/8/11/2019 financial crisis 2011 (1).pdf
11/16
the Federal Deposit Insurance Corporation extended its
insurance coverage to virtually all transactions deposits inOctober 2008, thereby eliminating depositors incentives
to pull funding from any transaction accounts. The effects
of these market pullbacks and policy steps can be seenclearly in the aggregate flows of deposits, graphed inFig. 2
Table 4
Fixed effects regressions of liquid asset, loan, and credit supply growth on crisis indicator, firm characteristics, and interactions.
This table reports fixed effects regressions of quarterly growth in liquid assets standardized by beginning of period assets. The table also reports fixed
effects regressions of growth in loans standardized by beginning of period assets and growth in credit supply (i.e., growth in loans plus growth in unused
commitments) standardized by beginning of period assets plus unused commitments. The data are observed quarterly for a panel of US commercial
banks over the period 2006Q1 through 2009Q2. Large banks are those banks with beginning of quarter assets greater than $1 billion, and small banks are
those banks with beginning of quarter assets less than $1 billion. Commercial bank data, aggregated at the high holding company as appropriate, are from
Call Reports of Condition and Income accessed via the Federal Reserve Bank of Chicago website (www.chicagofed.org). Banks with asset growth greater
than 10% during a quarter are dropped during that quarter to mitigate the effect of large mergers on changes in liquid assets, loans, and credit supply.
Variables are winsorized at the 1st and 99th percentiles. CRISISis an indicator variable that takes the value of one for observations that occur during theperiod 2007Q3 through 2009Q2 and zero otherwise. Standard errors, clustered at the firm level, are reported in parentheses. ***,**, and* denote that the
coefficients are statistically significantly different from zero at the 1%, 5%, and 10% level, respectively.
DLiquid Assetst/Assetst1 D(Cash US Treas Fed Funds Repost)/
Assetst1
DLoanst/
Assetst1
DCreditt/
(CommitAssets)t1
(1) (2) (3) (4)
Panel A: Large banks (total assets 4$1 billion)
Illiquid Assets/Assetst1 0.205*** 0.180*** 0.012 0.029*
(0.022) (0.019) (0.013) (0.017)
Illiquid/Assetst1nCRISIS 0.048*** 0.023** 0.021** 0.020**
(0.012) (0.009) (0.008) (0.009)
Core Deposits/Assetst1 0.002 0.000 0.022*** 0.041***
(0.012) (0.008) (0.008) (0.011)
Core Deposits/Assetst1nCRISIS 0.002 0.001 0.020*** 0.021***
(0.008) (0.007) (0.006) (0.007)
Capital/Assetst1 0.065 0.123*** 0.064* 0.109**
(0.048) (0.038) (0.035) (0.050)
Capital/Assetst1nCRISIS 0.033 0.010 0.063* 0.072**
(0.030) (0.026) (0.032) (0.033)
Commit/(CommitAssets)t1 0.069*** 0.064*** 0.189*** 0.020
(0.022) (0.018) (0.025) (0.031)
Commit/(CommitAssets)t1nCRISIS 0.020* 0.013 0.014 0.039***
(0.011) (0.010) (0.012) (0.011)
Log Assetst1 0.016*** 0.017*** 0.017*** 0.025***
(0.005) (0.004) (0.005) (0.007)
Log Assetst1nCRISIS 0.002** 0.001 0.002*** 0.002***
(0.001) (0.001) (0.001) (0.001)
Firm dummies Yes Yes Yes Yes
Quarterly time dummies Yes Yes Yes Yes
N 6,462 6,462 6,462 6,462
AdjustedR-squared 0.246 0.205 0.407 0.346
Panel B: Small banks (total assets o$1 billion)
Illiquid Assets/Assetst1 0.286*** 0.274*** 0.068*** 0.049***
(0.007) (0.007) (0.004) (0.005)
Illiquid/Assetst1nCRISIS 0.014*** 0.003 0.009*** 0.012***
(0.003) (0.003) (0.002) (0.002)
Core Deposits/Assetst1 0.032*** 0.044*** 0.005 0.001
(0.005) (0.005) (0.005) (0.006)
Core Deposits/Assetst1nCRISIS 0.003 0.008** 0.007** 0.022***
(0.004) (0.003) (0.003) (0.003)
Capital/Assetst1 0.016 0.015 0.073*** 0.071***
(0.011) (0.012) (0.009) (0.011)
Capital/Assetst1nCRISIS 0.006 0.029*** 0.017*** 0.021***
(0.006) (0.006) (0.004) (0.004)
Commit/(CommitAssets)t1 0.159*** 0.150*** 0.300*** 0.150***(0.010) (0.009) (0.009) (0.011)
Commit/(CommitAssets)t1nCRISIS 0.038*** 0.029*** 0.004 0.046***
(0.007) (0.007) (0.005) (0.007)
Log Assetst1 0.041*** 0.040*** 0.014*** 0.025***
(0.003) (0.003) (0.004) (0.005)
Log Assetst1nCRISIS 0.004*** 0.003*** 0.001*** 0.000
(0.000) (0.000) (0.000) (0.000)
Firm dummies Yes Yes Yes Yes
Quarterly time dummies Yes Yes Yes Yes
N 78,581 78,581 78,581 78,581
AdjustedR-squared 0.270 0.253 0.381 0.302
M.M. Cornett et al. / Journal of Financial Economics 101 (2011) 297312 307
http://www.chicagofed.org/http://www.chicagofed.org/8/11/2019 financial crisis 2011 (1).pdf
12/16
(which shows weekly changes in core and wholesale
deposits at commercial banks from September 10, 2008
through January 10, 2009).9 Wholesale deposits fell in
aggregate by almost $200 billion in the last quarter of
2008, while core deposits grew by about $500 billion in
aggregate. Given these flows, it should come as little
surprise that banks that were more reliant on core deposit
financing faced fewer liquidity problems during the crisis
than banks that relied more heavily on wholesale sources
of debt financing.
3.5. Magnitudes and macro-implications for credit
production
We offer two distinct ways to assess the economic
magnitude of liquidity exposure on credit production.
First, we reestimate our regressions for banks in different
size bins and report standardized coefficients, in which
both the dependent variable and each of the bank-level
characteristics are divided by the standard deviation of
that variable across the sample. (We do not normalize the
TED spread because this shock is common across the
sample.) We estimate our model separately for banks in
different size bins: those with assets below $100 million,
those with assets between $100 million and $500 million,
those with assets between $500 million and $1 billion,
and those with assets above $1 billion.
The results (presented inTable 5) show that liquidity
exposure mattered more in explaining how large banks
adjust credit growth to liquidity shocks, the TED spreadshock, than small banks. The standardized effects of the
TED interactions are largest for banks in the highest asset-
size bin in almost all cases. Thus, liquidity risk exposure
affects the adjustments of credit, relative to observed
variations across the sample, more for larger banks than
smaller banks. This seems, at first blush, counterintuitive
but could be understood by the fact that larger banks
entered the crisis much more exposed than smaller banks
across all four dimensions. They had lower capital, more
unused commitments, greater reliance on wholesale
funds, and higher holdings of illiquid assets (recall
Table 2).
Second,Table 6summarizes the economic magnitude,in dollar terms, implicit in our model. We answer the
Table 5
Fixed effects regressions of credit growth on TED spread, firm characteristics, and interactions: standardized regression coefficients.
This table reports fixed effects regressions of quarterly growth in credit supply (i.e., growth in loans plus growth in unused commitments),
standardized by beginning of period assets plus unused commitments. The data are observed quarterly for a panel of US commercial banks over the
period 2006Q1 through 2009Q2. Commercial bank data, aggregated at the high holding company as appropriate, are from Call Reports of Condition and
Income accessed via the Federal Reserve Bank of Chicago website (www.chicagofed.org). Banks with asset growth greater than 10% during a quarter are
dropped during that quarter to mitigate the effect of large mergers on changes in liquid assets, loans, and credit supply. Variables are winsorized at the
1st and 99th percentiles. TED spread is the quarterly average of the daily difference between the three-month London Interbank Offered Rate (LIBOR) and
the three-month US Treasury bill secondary market rate. LIBOR data are from the Bulgarian National Bank website ( www.bnb.bg/#) and Treasury bill
data are from the Federal Reserve Economic Data (FRED) website of the Federal Reserve Bank of St. Louis ( http://research.stlouisfed.org/fred2/). Standarderrors, clustered at the firm level, are reported in parentheses.***,**, and*denote that the coefficients are statistically significantly different from zero at
the 1%, 5%, and 10% level, respectively.
Large banks (total
assets 4$1 billion)
Medium banks (total
assets $500 million
to $1 billion)
Small banks (total
assets $100 million
to $500 million)
Smallest banks
(total assets
r$100 million)
(1) (2) (3) (4)
Illiquid Assets/Assetst1 0.116* 0.111** 0.180*** 0.383***
(0.068) (0.055) (0.028) (0.033)
Illiquid Assets/Assetst1nTED 5.253* 5.581*** 2.235** 1.493
(2.989) (2.146) (1.140) (1.187)
Core Deposits/Assetst1 0.230*** 0.117** 0.043* 0.037
(0.058) (0.048) (0.026) (0.034)
Core Deposits/Assetst1nTED 7.050** 6.608*** 4.565*** 3.929***
(2.886) (2.523) (1.045) (1.259)
Capital/Assetst1 0.134* 0.056 0.237*** 0.017
(0.082) (0.104) (0.040) (0.044)
Capital/Assetst1nTED 9.120** -0.612 2.349** 3.024*
(4.113) (2.315) (1.081) (1.605)
Commit/(CommitAssets)t1 0.048 0.148** 0.171*** 0.421***
(0.123) (0.070) (0.028) (0.028)
Commit/(CommitAssets)t1nTED 10.702*** 4.714* 5.057*** 2.295*
(3.322) (2.726) (1.287) (1.254)
Log Assetst1 1.246*** 0.325*** 0.641*** 0.708***
(0.318) (0.060) (0.048) (0.073)
Log Assetst1nTED 6.091* 1.751 0.078 1.573
(3.118) (2.091) (0.985) (1.121)
Firm dummies Yes Yes Yes Yes
Quarterly time dummies Yes Yes Yes Yes
N 6,462 7,097 38,166 33,318
AdjustedR-squared 0.342 0.347 0.300 0.367
9 Source is the Federal Reserves H8 weekly data on bank assets andliabilities.
M.M. Cornett et al. / Journal of Financial Economics 101 (2011) 297312308
http://www.chicagofed.org/http://www.bnb.bg/#http://research.stlouisfed.org/fred2/http://research.stlouisfed.org/fred2/http://www.bnb.bg/#http://www.chicagofed.org/8/11/2019 financial crisis 2011 (1).pdf
13/16
following hypothetical questions: How much less cash
and other liquid assets would banks have accumulated if
exposure to liquidity risk had been low throughout the
banking system? How much more credit would banks
have supplied had exposure to liquidity risk been low
throughout the system? (We assume in this exercise that
the coefficients would remain constant in an environment
of lower overall exposure to liquidity risk.)To answer these questions, in Panel A of Table 6 we
move each banks liquidity exposure to the lowest quar-
tile of the distribution and then reestimate the change in
credit (or liquid assets) implicit in our regression models
stemming from the TED spread shock observed in the
fourth quarter of 2008. For example, consider a bank with
$100 billion in assets, $60 billion in loans on balance sheet
and $20 billion in unused loan commitments as of the end
of 2007. Suppose that this hypothetical bank reduced its
stock of credit ($60 billion in loans $20 billion in unused
commitment) by 1% during the fall of 2008; that is,
suppose loans plus total unused commitments fell by
$800 million. These figures represent a large bank operat-ing at the 75th percentile of the distribution for the
commitment ratios (20%, recallTable 2). For such a bank,
we adjust its change in credit production as if it had
commitment exposure at the 25th percentile (11%) in the
face of the 250 basis points TED spread observed on
average during the fall of 2008. That is, for this bank we
estimate the following:
Actual change in credit $800 million,
Adjustment 0:110:200:0252:753$120 billion
$744 million,
and
Adjusted change in credit $56 million:
Theadjusted change in credit equals the actual change
in credit plus the product of the hypothetical movement
of liquidity exposure (0.110.20) times the TED spread
(250 basis points) times the estimated interaction term
(recallTable 3,Column 4) times the banks precrisis sum
of assets plus commitments ($120 billion). We then sum
up the adjusted change in creditacross all large banks to
arrive at an aggregate estimate of how much credit wouldhave changed in the fall of 2008 had all large banks
entered the quarter with low liquidity exposure. (We
make no adjustment for banks below the 25th percentile
of commitment exposure.) We estimate similar adjust-
ments across the other three dimensions of liquidity
exposure. That is, we move banks below the 75th per-
centile of the core deposits distribution up to the 75th
percentile; we move banks below the 75th percentile of
the capital-asset distribution to the 75th percentile; and
we move banks above the 25th percentile of the illiquid
assets distribution down to the 25th percentile. For each
of these changes, we aggregate up how liquid assets,
loans, and total credit would have changed in the fallof 2008.
Panel B ofTable 6reproduces a similar experiment but
uses the coefficients from the interaction of liquidity
exposure with the crisis indicator. These adjustments
are smaller and are perhaps a more conservative estimate
because the TED spread reached its maximum in
October 2008.
The total adjustments to liquidity accumulation and
credit production are very large.10 For example, large
Table 6
Economic impact of liquidity shocks during 2008Q4.
This table reports the estimated effect of the liquidity shock in the financial crisis on changes in liquid assets, loans, and credit supply in 2008Q4. We
adjust the actual changes in 2008Q4 as if each right-hand side variable observed below the 75th percentile (measured 2007Q4) had been equal to the
75th percentile level to simulate the response of the banking system if all banks had had low liquidity exposure. Panel A subjects banks to the 250 basis
points TED spread shock observed in 2008Q4 and uses coefficients from Table 3; Panel B subjects banks to the overall shock of moving into the financial
crisis regime and uses the coefficients fromTable 4.***,**, and* denote that the coefficient estimates are statistically significantly different from zero at
the 1%, 5%, and 10% levels, respectively. The data are observed quarterly for a panel of US commercial banks over the period 2006Q1 through 2009Q2.
DLiquid Assets2008Q4 D(Cash US Treas Fed Funds
Repos)2008Q4
DLoans2008Q4 DCredit2008Q4
(1) (2) (3) (4)
Panel A: Large banks (total assets 4$1 billion) (N 427), based on coefficients from Table 3
Actual change ($BN) 168 121 52 503
Illiquid Assets adjustment ($BN) 10*** 3 5* 6*
Core Deposits adjustment ($BN) 53 28 79*** 86**
Capital adjustment ($BN) 2 10 25** 26**
Commitadjustment ($BN) 101 65 249** 297***
Adjusted change ($BN) 2 15 192 87
Panel B: Large banks (total assets 4$1 billion) (N 427), based on coefficients from Table 4
Actual change ($BN) 168 121 52 503
Illiquid Assets adjustment ($BN) 8*** 4** 3** 3**
Core Deposits adjustment ($BN) 5 3 51*** 53***Capital adjustment ($BN) 6 2 11* 13**
Commitadjustment ($BN) 86* 56 60 168***
Adjusted change ($BN) 75 62 47 265
10 We estimate similar aggregates for small banks. These effects are
much smaller because most of the assets in the banking system are heldby large banks.
M.M. Cornett et al. / Journal of Financial Economics 101 (2011) 297312 309
8/11/2019 financial crisis 2011 (1).pdf
14/16
Table 7
Fixed effects regressions of credit supply growth on TED spread, firm characteristics, and interactions, with loan demand controls.
This table reports fixed effects regressions of quarterly growth in credit standardized by beginning of period credit. The data are observed quarterly for
a panel of US commercial banks over the period 2006Q1 through 2009Q2. Large banks are those banks with beginning of quarter assets greater than $1
billion, and small banks are those banks with beginning of quarter assets less than $1 billion. The first column replicates results from Table 3 for
comparison. Column 2 adds the share of commercial and industrial loans, the share of loans in real estate, and their interactions with TED to sweep out
potential demand effects. Column 3 instead introduces state fixed effects based on the banks headquarters and interactions between these state effects
and TED. Standard errors, clustered at the firm level, are reported in parentheses. ***, **, and * denote that the coefficients are statistically significantly
different from zero at the 1%, 5%, and 10% level, respectively.
Base model (Table 3, Column 4) With loan shares*TED With state*TED effects
(1) (2) (3)
Panel A: Large banks (total assets 4$1 billion)
Illiquid Assets/Assetst1 0.030* 0.027 0.038**
(0.017) (0.017) (0.017)
Illiquid Assets/Assetst1*TED 1.340* 0.897 1.787**
(0.762) (0.781) (0.791)
Core Deposits/Assetst1 0.044*** 0.049*** 0.045***
(0.011) (0.012) (0.012)
Core Deposits/Assetst1nTED 1.356** 1.594*** 0.940
(0.555) (0.567) (0.594)
Capital/Assetst1 0.088* 0.091* 0.095*
(0.054) (0.055) (0.055)
Capital/Assetst1nTED 5.984** 5.581** 5.093*
(2.699) (2.626) (2.847)
Commit/(CommitAssets)t1 0.012 0.009 0.019
(0.032) (0.031) (0.032)
Commit/(CommitAssets)t1nTED 2.753*** 3.676*** 2.613***
(0.855) (0.869) (1.006)
Log Assetst1 0.027*** 0.028*** 0.027***
(0.007) (0.007) (0.007)
Log Assetst1nTED 0.131* 0.112* 0.234***
(0.067) (0.067) (0.081)
Firm dummies Yes Yes Yes
Quarterly time dummies Yes Yes Yes
Loan share and loan sharesnTED No Yes No
State dummiesnTED No No Yes
N 6,462 6,462 6,462
AdjustedR-squared 0.342 0.345 0.347
Panel B: Small banks (total assets r$1 billion)
Illiquid Assets/Assetst1 0.051*** 0.058*** 0.049***
(0.006) (0.005) (0.006)
Illiquid Assets/Assetst1nTED 0.516*** 0.335* 0.504***
(0.183) (0.201) (0.187)
Core Deposits/Assetst1 0.002 0.008 0.001
(0.006) (0.006) (0.006)
Core Deposits/Assetst1nTED 1.553*** 1.394*** 1.155***
(0.238) (0.241) (0.242)
Capital/Assetst1 0.067*** 0.056*** 0.069***
(0.012) (0.010) (0.012)
Capital/Assetst1nTED 1.058*** 1.075*** 0.915***
(0.325) (0.343) (0.328)
Commit/(CommitAssets)t1 0.137*** 0.144*** 0.138***
(0.012) (0.012) (0.012)Commit/(CommitAssets)t1nTED 3.034*** 2.881*** 3.631***
(0.557) (0.553) (0.581)
Log Assetst1 0.028*** 0.038*** 0.027***
(0.006) (0.002) (0.006)
Log Assetst1nTED 0.021 0.030 0.063**
(0.024) (0.030) (0.026)
Firm dummies Yes Yes Yes
Quarterly time dummies Yes Yes Yes
Loan share and loan sharesnTED No Yes No
State dummiesnTED No No Yes
N 78,581 78,581 78,581
AdjustedR-squared 0.299 0.303 0.302
M.M. Cornett et al. / Journal of Financial Economics 101 (2011) 297312310
8/11/2019 financial crisis 2011 (1).pdf
15/16
banks accumulation of $168 billion in liquid assets during
2008Q4 falls to almost zero ($2 billion) after our adjust-
ment (Panel A). In other words, the model suggests that
there would have been no liquidity buildup in the face of
the 250 basis points TED spread had banks operated with
low levels of liquidity risk exposure going into the fall of
2008. Similarly, the drop in credit production would have
been nearly 90% lower had banks been less exposed. Theraw data indicate a drop in loans plus commitments of
$503 billion for the large banks. This decline drops to just
$87 billion after adjusting for liquidity risk. At the same
time, our adjustment shows that liquidity exposure
increased loans held on bank balance sheets. Total unad-
justed lending falls by $52 billion for large banks, whereas
the adjusted figure falls by $192 billion. This occurs
because firms drew on preexisting lines en masse during
the crisis.
The results are smaller but still substantial if we use
the coefficients from the crisis-indicator model (Panel B).
This model acts as if the shocks to liquidity were the same
in all quarters from the middle of 2007 on and thusunderstates the impact during 2008Q4, when TED
reached its apogee and credit declined most dramatically.
Nevertheless, the adjustments to the predicted decline in
credit would have reduced the decline nearly 50% (from
$503 billion down to $265 billion).
The aggregation also highlights the economic impor-
tance of both core deposits as a stabilizing source of funds
and undrawn commitments as a major source of destabi-
lizing asset-side liquidity exposure. If large banks had
held core deposits at the 75th percentile or higher, our
calculation suggests that credit production would have
been higher by $86 billion, and if all banks had exposure
to undrawn commitments at the 25th percentile credit orlower, credit would have grown by $297 billion more.
These two effects dominate the aggregates both because
the estimated coefficients are large and because large
banks were much more exposed to liquidity risk going
into the crisis. The relative importance of core deposits in
the funding structure of banks tends to decrease with size,
while the relative importance of undrawn commitments
increases with size. Thus, the adjustments to credit tend
to be larger for large banks than for small banks.
3.6. Robustness tests
Loan demand probably began to decline during thecrisis quarters and, thus, could play some role in explain-
ing the drop in credit production. Because our model
includes bank fixed effects and time indicators, and
because we focus only on the interaction between TED
and liquidity exposure, demand explanations could drive
our interaction effects only if two conditions hold: (1)
loan demand must be correlated with within-bank varia-
tion in our measures of liquidity risk and (2) loan demand
must fall more at banks with high liquidity risk when the
economy moves from boom (low TED spread) to bust
(high TED spread) than at banks with low liquidity risk.
Table 7shows that adding variables plausibly related to
changes in demand conditions had little impact on ourfindings. In the first approach, we control for differences
in loan shares across banks, and in the second we control
for differences in geographical markets.
To be specific, we estimate two robustness tests to
sweep out possibly omitted demand factors. First, we
introduce the share of real estate loans to total loans and
the share of business loans to total loans as right-hand-
side regressors, along with interactions between each of
these with the TED spread. Second, we sweep out poten-tial demand variation related to geographical location of
borrowers by adding a set of state-level indicator vari-
ables and their interaction with the TED spread. For a
given bank, we define a state indicator to equal one if the
bank has branches located in that state, based on the
branch-level data at the FDICs Summary of Deposits.
Because most business lending, particularly lending to
small business, relies on monitoring facilitated by close
geographic proximity, branch location correlates closely
with borrower location. For example, Berger, Miller,
Petersen, Rajan, and Stein (2005) report a median
distance between small borrowers and their bank of
just three miles. Average distance does increase, however,with bank size. Large banks are more likely to lend
using information technology such as credit scoring as a
substitute for personal connections with borrowers.
Thus, this second robustness check probably works very
well for small banks but could be less effective for
large banks.
The results of these robustness tests are reported in
Table 7. Because the emphasis here is on loan demand
variation, we report only the models of total credit
production. The first column of Table 7 reproduces the
baseline models fromTable 3, Column 4. These data show
that our results of interest are stable even when we
introduce two distinct approaches to sweep out demand.In every case, the interaction between TED and the
liquidity variables maintain similar sign and magnitude.
We lose little statistical significance. No evidence exists
that coefficients are systematically moving toward or
away from zero (e.g., no evidence of attenuation bias or
evidence that we are overstating the effects of liquidity
exposure). In some cases, coefficients increase slightly in
magnitude, while in others they decline slightly. In no
cases, however, do the effects change much relative to
sampling error.
4. Conclusions
Liquidity at banks dried up during the financial crisis of
20072009, both because interbank markets froze and
because markets for asset-backed and mortgage-backed
securities collapsed. Illiquidity peaked in the fourth quarter
of 2008 after the failure of Lehman Brothers and the AIG
bailout. The Fed first attempted to stabilize the financial
system with traditional tools of monetary policy, then
expanded their balance sheet by more than $1 trillion over
a few weeks, and later implemented new techniques such
as equity injections and extensions of liability guarantees. In
this paper, we show how this expansion of liquidity was
distributed across the banking system. We find that banks
with more illiquid asset portfolios, i.e., those banks that heldmore loans and securitized assets, increased their holdings
M.M. Cornett et al. / Journal of Financial Economics 101 (2011) 297312 311
8/11/2019 financial crisis 2011 (1).pdf
16/16
of liquid assets and decreased lending. We also find that
banks that relied more heavily on stable sources of finan-
cing, i.e., core deposits and capital, continued to lend relative
to other banks. Off-balance sheet liquidity risk, in the form
of undrawn loan commitments, materialized as borrowers
drew on preexisting commitments in large quantities. These
takedowns displaced lending capacity and constrained new
credit origination. When we aggregate our results up, wefind that most of the decline in bank credit production
during the height of the crisis can be explained by liquidity
risk exposure.
References
Acharya, V.V., Schnabl, P., 2010. Do global banks spread global imbal-ances? The case of asset-backed commercial paper during thefinancial crisis of 200709. IMF Economic Review 58, 3773.
Ashcraft, A.B., 2008. Does the market discipline banks? New evidencefrom the regulatory capital mix. Journal of Financial Intermediation17, 543561.
Avery, R.B., Belton, T.M., Goldberg, M.A., 1988. Market discipline inregulating bank risk: new evidence from the capital markets. Journal
of Money, Credit, and Banking 20, 597610.Berger, A.N., Bouwman, C.H.S., 2009. Bank liquidity creation. Review of
Financial Studies 22, 37793837.Berger, A. N., Bouwman, C. H. S., 2010. How does capital affect bank
performance during financial crises? SSRN: /http://ssrn.com/abstract=1739089S.
Berger, A., Miller, N., Petersen, M., Rajan, R., Stein, J., 2005. Does functionfollow form? Evidence from the lending practices of large and smallbanks. Journal of Financial Economics 76, 237269.
Bhattacharya, S., Thakor, A.V., 1993. Contemporary banking theory.Journal of Financial Intermediation 3, 250.
Billett, M., Garfinkel, J., ONeal, E., 1998. The cost of market vs. regulatorydiscipline. Journal of Financial Economics 48, 333358.
Black, H.A., Collins, M.C., Robinson, B.L., Schweitzer, R.L., 1997. Changesin market perception of riskiness: the case of too-big-to-fail. Journalof Financial Services Research 20, 389406.
Brunnermeier, M.K., 2009. Deciphering the liquidity and credit crunch
20072008. Journal of Economic Perspectives 23, 77100.Diamond, D.W., Dybvig, P.H., 1983. Bank runs, deposit insurance, and
liquidity. Journal of Political Economy 91, 401419.Diamond, D.W., Rajan, R.G., 2000. A theory of bank capital. Journal of
Finance 55, 24312465.
Diamond, D.W., Rajan, R.G., 2001a. Banks and liquidity. American
Economic Review 91, 422425.Diamond, D.W., Rajan, R.G., 2001b. Liquidity risk, liquidity creation, and
financial fragility: a theory of banking. Journal of Political Economy
109, 287327.Flannery, M.J., 1998. Using market information in prudential bank
supervision: a review of the US empirical evidence. Journal ofMoney, Credit, and Banking 30, 273305.
Flannery, M.J., 2001. The faces of market discipline. Journal of Financial
Services Research 20, 107119.Gatev, E., Schuermann, T., Strahan, P.E., 2009. Managing bank liquidity
risk: how depositloan synergies vary with market conditions.Review of Financial Studies 22, 9951020.
Gatev, E., Strahan, P.E., 2006. Banks advantage in hedging liquidity risk:
theory and evidence from the commercial paper market. Journal ofFinance 61, 867892.
Gorton, G., 2009. Slapped in the face by the invisible hand: banking andthe panic of 2007. SSRN: /http://ssrn.com/abstract=1401882S.
Gorton, G. Metrick, A., 2009. Securitized banking and the run on therepo. Unpublished Working Paper 15223. National Bureau of Eco-nomic Research, Cambridge, MA.
Gorton, G., Pennacchi, G., 1990. Financial intermediaries and liquiditycreation. Journal of Finance 45, 4971.
Gorton, G. B., Winton, A., 2000. Liquidity provision, bank capital, and themacroeconomy. SSRN: /http://ssrn.com/abstract=253849S.
Goyal, V.K., 2005. Market discipline of bank risk: evidence from sub-
ordinated debt contracts. Journal of Financial Intermediation 14,318350.
Hannan, T.H., Hanweck, G.A., 1988. Bank insolvency risk and the marketfor large certificates of deposit. Journal of Money, Credit, and Bank-ing 20, 203211.
Houston, J., James, C., Marcus, D., 1997. Capital market frictions and therole of internal capital markets in banking. Journal of Financial
Economics 46, 135164.Ivashina, V., Scharfstein, D., 2010. Loan syndication and credit cycles.
American Economic Review 100, 5761.Kashyap, A.K., Rajan, R., Stein, J.C., 2002. Banks as liquidity providers: an
explanation for the coexistence of lending and deposit taking.
Journal of Finance 57, 3373.Loutskina, E., The role of securitization in bank liquidity and funding
management. Journal of Financial Economics, in press, doi: 10.1016/j.
jfineco.2011.02.005.Maechler, A.M., McDill, K.M., 2006. Dynamic depositor discipline in U.S.
banks. Journal of Banking and Finance 30, 18711898.OHara, M., Shaw, W., 1990. Deposit insurance and wealth effects: the
value of being too big to fail. Journal of Finance 45, 15871600.Pennacchi, G.G., 2009. Deposit insurance. Unpublished Working Paper,
University of Illinois, UrbanaChampaign.
M.M. Cornett et al. / Journal of Financial Economics 101 (2011) 297312312
http://ssrn.com/abstract=1739089http://ssrn.com/abstract=1739089http://ssrn.com/abstract=1401882http://ssrn.com/abstract=253849http://localhost/var/www/apps/conversion/tmp/scratch_10/dx.doi.org/10.1016/j.jfineco.2011.02.005http://localhost/var/www/apps/conversion/tmp/scratch_10/dx.doi.org/10.1016/j.jfineco.2011.02.005http://localhost/var/www/apps/conversion/tmp/scratch_10/dx.doi.org/10.1016/j.jfineco.2011.02.005http://localhost/var/www/apps/conversion/tmp/scratch_10/dx.doi.org/10.1016/j.jfineco.2011.02.005http://ssrn.com/abstract=253849http://ssrn.com/abstract=1401882http://ssrn.com/abstract=1739089http://ssrn.com/abstract=1739089